Copyrieht 2001 by the American Psychological Association. Inc. 0096-1523/01/55.00 DO!: 10.1037//0096-1523.273.1243
Journal of Experimental Psychology: Human Perception and Performance 2001 Vol. 27. No. 5. 1243-1259
Coupling of Breathing and Movement During Manual Wheelchair Propulsion Polemnia G. Amazeen, Eric L. Amazeen, and Peter J. Beek Vrije Universiteit, Amsterdam The hypothesis of this study was that stable coordination patterns may be found both within and between physiological subsystems. Many studies have been conducted on both monofrequency and multifrequency coordination, with a focus on both the frequency and phase relations among the limbs. In the present study, locomotor-respiratory coupling was observed in the maintenance of small-integer frequency ratios (2:1, 3:1, and 4:1) and in the consistent placement of the inspiratory phase just after the onset of the movement cycle during wheelchair propulsion. Level of experience and various motor and respiratory parameters were manipulated. Coupling was observed across levels of experience. Increases in movement frequency were accompanied by a shift to larger-integer ratios, suggesting that a single modeling strategy (e.g., the Farey tree; D. L. Gonzalez & O. Piro, 1985) may be used for coordination both within the motor subsystem and between it and other physiological subsystems.
Humans and animals alike demonstrate a discrete number of stable coordination patterns. Patterns of locomotion, called gaits, are limited in number and easily recognizable. For quadrupeds, the three most common gaits are the walk, trot, and gallop, although more complex subdivisions of gait are possible (for an overview, see Collins & Stewart, 1993; Schoner, Jiang, & Kelso, 1990). In bipedal locomotion, spontaneously produced patterns are limited to the walk or run (antiphase) and the jump (inphase) patterns. In all of these patterns, the limbs move at the same frequency, so that the different patterns may be characterized by different phase relations between the limbs. When the constraint of postural stability is removed, observed motor patterns become more complex and different phase relations (e.g., 90°, Zanone & Kelso, 1992, 1997) can be acquired. The HKB model, a dynamical model first developed by Haken, Kelso, and Bunz (1985), accommodates the different monofrequency coordination patterns observed including the effects of learning, handedness, and attention (see summary in Amazeen, Amazeen, & Turvey, 1998). Multifrequency coordination patterns—in which the limbs move at different frequencies, as in drumming (e.g., Peper, Beek, & van Wieringen, 1995a, 1995b), and may be spontaneously produced or learned—are also accom-
modated by an expansion of the HKB model (Sternad, Turvey, & Saltzman, 1999a, 1999b) or by other mathematical structures like the Farey tree (e.g., Gonzalez & Piro, 1985). In all of the aforementioned patterns, coordination takes place between two or more limbs or limb segments, that is, within a motor subsystem. The present study was designed to explore the possibility that coordination may take place between physiological subsystems. The presence of intermodal coupling would suggest that coordination is a basic phenomenon that may be observed whenever two or more rhythmic processes interact. The phenomenon we focused on was locomotor-respiratory coupling (LRC), the pacing of breathing while exercising. If LRC is observed, then the observed patterns may indicate some common basis for modeling.
Locomotor-Respiratory Coupling Coupling may be operationalized as either the consistent phasing (e.g., inphase) or maintenance of a small-integer frequency relation (e.g., 2:1) of two rhythmic processes. In this sense, the term applies to coordination within the motor subsystem, where the focus is on phasing during monofrequency coordination and on frequency ratios during multifrequency coordination. Because the movements of a limb or limb segment are usually faster than respiration, LRC is comparable to multifrequency coordination; therefore, the literature focuses predominantly (but not exclusively) on identifying the frequency ratios that are used. LRC has been observed during both quadrupedal and bipedal locomotion. LRC was first studied in a young jackrabbit that was trained to run on a treadmill (see Bramble & Carrier, 1983). At low speeds, it completed two full breathing cycles per locomotory cycle. At higher speeds, it switched to a 1:1 (monofrequency) ratio between stride frequency and respiratory frequency. The components of LRC were studied more extensively in another quadruped, the horse, which maintains a constant 1:1 ratio during the canter (slow gallop) and gallop (Bramble & Carrier, 1983; Lafortuna, Reinach, & Saibene, 1996). Human runners demonstrate more flexibility than quadrupeds, shifting from 4:1 (four strides per breath) to 2:1 with
Polemnia G. Amazeen, Eric L. Amazeen, and Peter J. Beek, Faculty of Human Movement Sciences, Vrije Universiteit, Amsterdam, the Netherlands. Eric L. Amazeen is now at the Department of Psychology, Arizona State University. Preliminary findings were reported in Amazeen, Amazeen, and Beek (1999). Support was provided by Vrije Universiteit Grant USF'96 awarded to Peter J. Beek. We would like to acknowledge the assistance of Luc van der Woude, Annet Dallmeijer, Jos van den Berg, and Bert Clairbois in the preparation of the experiments and the assistance of Jay Holden in the comparison of left and right limb movements. Correspondence concerning this article should be addressed to Polemnia G. Amazeen, who is now at the Department of Psychology, Arizona State University, P.O. Box 871104, Tempe, Arizona 85287-1104. Electronic mail may be sent to
[email protected]. 1243
AMAZEEN, AMAZEEN, AND BEEK
1244
increased velocity and altering frequency ratios (e.g., 2:1,3:1,4:1,3:2, 5:2) during steady-state behavior (Bramble & Carrier, 1983). Phase analyses that have been conducted on human runners reveal a preference for initiating and terminating a respiratory cycle with either the left or right foot during level track running when even ratios (e.g., 2:1, 4:1) are used (Bramble & Carrier, 1983). Phase relations differ for uphill and downhill running; inspiration was observed during the support phase of the step in uphill running and during the swing phase of the step in downhill running (Takano, 1995). No consistent phasing was observed in relation to the onset of expiration, which suggests that inspiration is more closely coupled to motor processes than is expiration. Individual differences—that is, consistency of phasing strategy within an individual but not between individuals—have been documented as well (Bernasconi & Kohl, 1993), although the development of those individual strategies has not been studied. When Is LRC Observed? A large number of studies have been conducted on LRC during human lower-limb locomotion, like running and bicycling. Untrained bicyclists have been shown to maintain a 2:1, 3:1,4:1, 6:1, 3:2, and 5:2 ratio of pedal to respiratory frequency (Garlando, Kohl, Koller, & Pietsch, 1985; Paterson. Wood, Morton, & Henstridge, 1986). The variability of ratios observed may indicate that LRC is more stable in some activities than in others. A direct comparison of running and cycling revealed a greater degree of coupling, as defined by phase, during running (Bernasconi & Kohl, 1993). In general, experienced runners and cyclists have been shown to have a higher degree of coupling than untrained individuals (Bramble & Carrier, 1983; Kohl, Koller, & Jager, 1981), although no systematic study of LRC changes during training in these sports has been conducted. Although it is tempting to believe that LRC only occurs when both frequency and phase coupling are observed, an absence of frequency and/or phase locking does not necessarily imply that coupling is absent. According to the theory of coupled oscillators, coupling between two rhythmic processes may be too weak to induce both frequency and phase locking. Evidence of weaker coupling takes the form of relative coordination, in which the component processes pass into and out of a coordinated state rather than maintain coordination steadily (von Hoist, 1939). Relative coordination may be practically observed as bouts of coupling and decoupling or as switching among frequency ratios and/or phase relations. Although it is tempting to think of weak coupling as synonymous with inexperience, a notable observation is the intentional use of relative coordination by competitive cyclists, who report that the intentional decoupling of locomotion and respiration allows them to shift gear ratios more easily (Garlando et al., 1985). The uncoupled state cannot be maintained because oxygen consumption is least—and therefore exercise is most efficient— during LRC (Bernasconi & Kohl, 1993; Garlando et al., 1985). Therefore, experience may not provide for stronger coupling but rather for greater control over the presence or absence of coupling. A number of studies have looked for a coupling mechanism to derive a rule for the presence or absence of coupling. The seemingly greater efficiency of coupling over noncoupling seems to imply a bidirectional influence, and yet LRC is largely assumed to be unidirectional, from the motor subsystem to the respiratory
subsystem, rather than vice versa (Astrand & Rodahl, 1977; Bechbache & Duffin, 1977; Bramble & Carrier, 1983; Kao, 1963; Paterson et al., 1986). One of the most widely accepted mechanisms for LRC has been the visceral piston, a metaphor for the rhythmic perturbation of the diaphragm by any vertical impulse (Bramble & Carrier, 1983; Bramble & Jenkins, 1993; Paterson et al., 1986). This model assigns causal priority to the motor subsystem rather than to respiration. During running, vertical impulses are generated by the ground reaction force of footfalls, producing the coupling that is observed between movements and respiration. In contrast, any activity that involves only the upper limbs does not generate a vertical impulse and so, according to the visceral piston hypothesis, should not produce LRC. Evidence from the literature on upper-limb locomotion suggests otherwise. LRC has been observed in bats and various species of flying birds (e.g., Butler & Woakes, 1980; Suthers, Thomas, & Suthers, 1972) and in the human activity of rowing (Mahler, Hunter, Lentine, & Ward, 1991; Mahler, Shuhart, Brew, & Stukel, 1991). In one study, elite rowers demonstrated frequency ratios of 1:1 and 1:2 (rowing strokes per breath) and inspired at phases of the rowing cycle that were consistent within each rower but differed across rowers (Mahler, Shuhart, et al., 1991), suggesting different LRC strategies. In another study, novices produced the 1:2 ratio only during submaximal exercise intensities (Mahler, Hunter, et al., 1991). Following training, they became much more consistent, maintaining a 1:2 ratio during both submaximal and peak exercise and developing a phasing strategy in which they inspired just after the beginning and end of each rowing stroke. The lack of a vertical impulse in rowing implies that the visceral piston is not solely responsible for producing LRC. Rather, LRC may be observed whenever there is a simple mechanical interaction between the motor and respiratory subsystems. The evidence from the motor coordination literature—particularly the finding that the HKB model holds for coordination both within a single individual and between individuals (e.g., Amazeen, Schmidt, & Turvey, 1995; Schmidt, Carello, & Turvey, 1990)—suggests that some form of interaction between two systems is all that is required for stable coordination to occur. Although rowing is not a prototypical form of locomotion, wheelchair propulsion is for many disabled populations. This leads to a number of practical reasons for its study, including the development of efficient rehabilitation strategies for wheelchair users. In the only study to date on LRC during manual wheelchair propulsion, wheelchair basketball players were asked to propel a wheelchair on a treadmill with a 3% slope at speeds below, at, and above their previously established comfortable wheelchair speeds (MacDonald, Kirby, Nugent, & MacLeod, 1992). A frequencybased definition of LRC was used in which coupling was defined as the coincidence of characteristic frequencies from the power spectra of the propulsion and respiratory time series. A frequency ratio of 3:1 was witnessed on 40% of the trials, although breathby-breath analyses revealed that the coupling ratio for each breath alternated among 2:1, 3:1, and 4:1 patterns. Both propelling and respiratory frequency increased as a function of velocity, but the ratio of the two was indifferent to the velocity manipulation. A control analysis was performed by calculating the ratios across the propulsion and respiratory frequencies of different participants. MacDonald et al. (1992) concluded that LRC had not actually occurred because the incidence of "false" coupling was 27%. They
LOCOMOTOR-RESPIRATORY COUPLING
attributed the absence of LRC during wheelchair propulsion to the exclusion of the abdominal visceral piston. However, given that LRC occurs during rowing (Mahler, Hunter, et al., 1991; Mahler, Shuhart, et al., 1991), it is equally plausible that the methodological constraints of MacDonald et al.'s experiment precluded the observation of LRC during manual wheelchair propulsion. Goals of the Present Study The present set of experiments was designed to reconsider the occurrence of LRC during manual wheelchair propulsion. LRC clearly occurs during lower-limb activities like running and bicycling. Observations of frequency and phase coupling during upperlimb locomotion will strengthen the argument that coordination is a general phenomenon that occurs both between and within bodily subsystems. Three experiments were conducted that addressed (a) the existence of coupling during upper-limb locomotion, (b) relevant parameters that cause changes in frequency ratios, and (c) the directionality of the coupling. If patterns of LRC mimicked the results of other tasks limited to interlimb coordination, then we expected that a common basis for modeling could be discovered. The precise form of the modeling would depend on the particular pattern of results that is observed. To improve the likelihood of observing LRC, we made a number of methodological alterations to the MacDonald et al. (1992) study. First, we tested able-bodied participants, so that we were able to avoid the influence on LRC of potential respiratory and/or motor problems that are associated with spinal cord injuries or that may have resulted from secondary medical complications associated with wheelchair use (Glaser, Sawka, Young, & Suryaprasad, 1980; Janssen, van Oers, Hollander, Veeger, & van der Woude, 1993). Nonwheelchair-dependent populations also tend to be more homogeneous and can be challenged to a greater degree during wheelchair exercises (van der Woude, van Croonenborg, Wolff, Dallmeijer, & Hollander, 1999). Challenge may be operationalized as the participant's energy expenditure or power output, which is the product of velocity and any resistance (e.g., air and friction) that the participant encounters during the task. MacDonald et al. (1992) tested the impact of velocity only on LRC, but other experiments on wheelchair propulsion have varied the rolling resistance of the wheels to control for resistance and, therefore, to influence power output (e.g., van der Woude et al., 1988). Therefore, a second alteration to MacDonald et al.'s methodology that we introduced was the manipulation of both velocity and rolling resistance in Experiment 1, to determine both their individual effects and their joint influence (through power output) on LRC. Movement frequency was manipulated in Experiment 2 to provide for comparisons of observed frequency ratios with the multifrequency coordination literature. Two alterations were made to the analyses that were performed by MacDonald et al. (1992). First, because participants are likely to alter their propulsion and/or respiratory patterns during the course of a trial, cycle-by-cycle analyses of frequency ratio were performed instead of using summary values. A control analysis of the kind performed by MacDonald et al. was not used because individual participants with similar physical abilities are likely to elect the same characteristic frequency of propulsion for similar task constraints. For example, in the present study, 2 male participants elected a propulsion frequency of 0.97 Hz when asked to propel the wheelchair ergometer at 3 km/hr. Even if their respira-
1245
tory frequencies differed to produce a conclusion of LRC within each individual (e.g., 0.32 Hz for Participant A yields a ratio of 3:1, and 0.24 Hz for Participant B yields a ratio of 4:1), calculating the ratios across participants (yielding 4:1 and 3:1) would have negated the LRC conclusion. Second, both frequency and phasing analyses were performed to determine the degree, and possibly the directionality, of the coupling. A conclusion of coupling—as indexed by the occurrence of frequency and/or phase locking— would reinforce the notion of a necessary coordination of the motor and respiratory subsystems during upper-limb activity. Experiment 1: Demonstration of LRC The first experiment was simply directed at the demonstration of LRC during manual wheelchair propulsion. Able-bodied participants were tested to eliminate the possible confounding influence of respiratory and/or motor dysfunction on coupling. If LRC occurs, then it is possible that able-bodied wheelchair users, like runners, may alter their frequency ratios as a function of velocity. A unique aspect of wheelchair propulsion is that power output may be manipulated by varying both velocity and the rolling resistance of the wheels. Both manipulations were performed in the present experiment to determine whether frequency ratios and/or phasing would change as a function of the relative challenge of wheelchair propulsion to the motor subsystem.
Method Participants. Seven able-bodied participants (5 men, 7 women; 25-44 years old; all right-handed) volunteered to participate in the experiment. Their wheelchair experience ranged from never having propelled a wheelchair (n = 5) to having participated in and conducted wheelchair experiments for 5 (n = 1) to 15 (n = 1) years. Participants were asked to refrain from smoking and from ingesting caffeine and/or alcohol for at least 2 hr prior to testing. They were asked to eat a light meal 2 hr before the experiment. All participants were naive to the purpose of the experiment. To retain the experience level of each of the participants, we did not permit any of the participants to practice with the equipment prior to the experiment. Apparatus. The experiment was run on a stationary wheelchair ergometer (Figure 1), whose physical dimensions and rolling characteristics could be altered to accommodate the anatomical dimensions and physical abilities of the individual participant (Neising et al., 1990). Individual adjustment of the seat height and width was required to maximize the efficiency of wheelchair propulsion. Following experimental protocol of past studies (e.g., van der Woude et al., 1999; Veeger, Lute, Roeleveld, & van der Woude, 1992), we adjusted the seat height until the participant's elbow angles were 110° (with 180° defined as full extension), and we adjusted the seat width until either the participant's shoulders were located directly over the wheel rims or, if the shoulders were not wide enough, the sides of the wheelchair seat rested against the participant's hips. The seat position was recorded and maintained during the course of the experiment. A monitor that was placed at eye height in front of the participant displayed both the required and actual velocity of the wheels. Propulsion frequency was indexed as the rate of torque application to the right wheel rim,1 which was measured directly from the ergometer at 100 Hz. An air flow measurement system, or pneumotachometer (Hans Rudolph, Inc., Kansas City, MO) was used to measure respiratory frequency by means of the differential pressure method. Two membranes that were
1
The correlation between the right and left wheel rim exceeded .90 for all dependent measures.
AMAZEEN, AMAZEEN, AND BEEK
1246
\ beep/ t%
•?*•
Match Required Velocity
Monitor
Computer Figure 1. Experimental arrangement for the study of locomotor-respiratory coupling during manual wheelchair propulsion. From "Locomotor-Respiratory Coupling During Manual Wheelchair Propulsion," by P. G. Amazeen, E. L. Amazeen, and P. J. Beck, 1999. In L. H. V. van der Woude, M. T. E. Hopman, and C. H. van Kemenade (Eds.), Ergonomics of Manual Wheelchair Propulsion: State of the Art II (p. 203), Amsterdam, the Netherlands: IOS Press. Copyright 1999 by IOS Press. Reprinted with permission.
mounted in a mouthpiece registered the flow at 100 Hz. The mouthpiece was attached to a headband that the participant wore, and a nose clip was used to guarantee oral respiration. The propulsion and respiratory time series were synchronized. Analyses were conducted to identify the onset of respiratory and propulsion cycles as sharp, positive accelerations of the respiratory and propulsion time series, respectively. (Raw data are presented in Figure 2.) Algorithms. An 18-point running average was computed to smooth both the respiratory and propulsion time series. The first derivative of these smoothed time series was calculated, and maximal accelerations were identified as the onset of inspiration and the onset of propulsion, respectively. Respiratory and propulsion frequencies were calculated cycle-bycycle as the sampling frequency (100 Hz in Experiment 1; 50 Hz in Experiments 2 and 3) divided by the difference between one onset and the preceding cycle's onset. Each of these frequency time series was expanded to a time series of 12,000 points so that we could determine both the frequency ratio and relative phase. In addition, we represented the propulsion phase with a number between 0° and 360°, with 0° set as the onset and 360° set as the termination of each propulsion cycle. Both frequency ratio and relative phase were calculated using the onset of inspiration as a landmark. Effectively, then, there were as many frequency ratio and relative phase calculations as there were respiratory cycles for each trial. Frequency ratio was calculated by dividing the propulsion frequency by the respiratory frequency at the onset of inspiration. Relative phase was defined as the propulsion phase minus the respiratory phase. We identified this as the phase of propulsion at the onset of inspiration. On the basis of this convention, a positive relative phase value was indicative of the propulsion cycle being initiated prior to inspiration, and a negative relative phase value was indicative of the propulsion cycle being delayed until after the onset of inspiration. Procedure. Prior to the experiment, each participant's maximum load (ML) was determined so that manipulations of the wheelchair's rolling resistance (in newton meters) could be scaled to individual abilities. ML was operationally defined as the maximum rolling resistance for which a participant could sustain a maximum velocity of 6.5 km/hr (the maximum velocity used in the experiment) for 5 min. ML varied widely across participants (26-40 Nm across both left and right wheels). Participants were also given an opportunity, on a separate trial, to adjust to the mouthpiece by breathing through it for a period of 5 min. Once the experiment began, the mouthpiece was worn during each 5-min trial. On any given trial, each participant sustained a constant velocity of 3.6, 5.0, or 6.5 km/hr at one of three loads (10 Nm, 0.5 ML, or 1.0 ML). The two lower
velocities were within the range of velocities used in previous experiments on wheelchair propulsion in both spinal cord injured and able-bodied participants (e.g., Dallmeijer, van der Woude, Hollander, & Angenot, 1999; van der Woude et al., 1999; van der Woude, Veeger, Rozendal, & Sargeant, 1989); we included the highest velocity condition to test the possibility that LRC might occur only during extreme testing conditions. Because of technical limitations on the total sample size, data were collected at 100 Hz during the last 40 s of the 2nd, 3rd, 4th, and 5th min of each trial, yielding propulsion and respiratory time series of 4,000 data points each. Trial order was randomized and participants were permitted to rest after each trial for at least 5 min. The experiment was conducted in two 1-hr sessions on 2 different days so that we could avoid the effects of fatigue. All procedures reported in the present experiments adhere to the ethical guidelines of the American Psychological Association. They were approved by the Research Committee of the Faculty of Human Movement Sciences of the Vrije Universiteit, Amsterdam.
Results Figure 2 depicts the raw, unsmoothed time series for two representative participants. Note that each point of inhalation, which marks the onset of the respiratory cycle, is accompanied by a push to the right wheel rim. There are consistently two propulsion cycles per respiratory cycle (i.e., a 2:1 frequency ratio) in Figure 2A and four propulsion cycles per respiratory cycle (4:1) in Figure 2B. In both instances, the onset of the respiratory cycle is temporally synchronized with the onset of the propulsion cycle. This inphase relation is depicted in Figure 3 for the 2 more experienced and the 5 less experienced (novice) participants. Relative phase is displaced slightly in the positive direction for novices, revealing a preference for initiating the propulsion cycle slightly prior to inhaling, and in the negative direction for more experienced wheelchair users, indicating that they tended to delay the propulsion cycle until the onset of inspiration. The reliability of this observation was tested when experience was manipulated explicitly in Experiment 2. The ratio of propulsion to respiratory frequency was calculated for every respiratory cycle individually. Because the number of respiratory cycles differed widely across individuals (2-20 for 40 s), the frequency distributions of Figures 4A-4H depict the percentage of the total number of cycles that were spent perform-
LOCOMOTOR-RESPIRATORY COUPLING
the peaks accounts for more than 40% of the cycles performed, although combining all peaks accounts for an average of at least 50% of performance. Given that experience appears to be a relevant variable, we combined individual performance post hoc according to level of experience in Figure 4H. Note that although LRC appears to occur for both groups, performance is more clearly defined and less variable for more experienced wheelchair users. The reliability of this observation was tested when experience was manipulated explicitly in Experiment 2. An analysis of variance (ANOVA) was performed to determine the effect of velocity and rolling resistance on respiratory frequency, propulsion frequency, and the ratio of the two. Although both respiratory frequency (M = 0.264, 0.288, and 0.328 Hz for 3.6, 5.0, and 6.5 km/hr, respectively) and propulsion frequency (M = 0.684, 0.760, and 0.858 Hz for 3.6, 5.0, and 6.5 km/hr, respectively) increased significantly with increased velocity, F(2, 12) = 8.73,p < .005 and F(2,12) = 11.70, p < .005, respectively, the cycle-by-cycle frequency ratio was unaffected by the velocity manipulation, F(2, 12) < 1. This replicates MacDonald et al.'s (1992) finding of a lack of influence of velocity on LRC. An increase in rolling resistance was accompanied by an increase in only the frequency of propulsion (M = 0.708, 0.750, and 0.836 Hz for 10 Nm, 0.5 ML, and 1.0 ML, respectively), F(2, 12) = 5.04, p < .005, and did not affect the cycle-by-cycle frequency ratio, F(2, 12) < 1. Therefore, although manipulations of both velocity and rolling resistance bring about changes in other relevant measures in wheelchair propulsion (e.g., gross mechanical efficiency; van der Woude et al., 1988), they do not alter the LRC pattern.
LJJ D
B
=
I = LU Q 3
li
•
1247
E
: = = I
i
0.
